Human Responses and LLM-Generated Follow-Up Dialogue for a 54-Item Questionnaire

This dataset contains multistage interview responses (in Russian) collected from 10 human participants who answered 54 primary questions covering life history, personal values, interests, work, hobbies, and future aspirations. Each initial response was evaluated by an LLM-based interviewer agent, which generated a follow-up question whenever the original answer was incomplete or underspecified. Follow-up responses were then produced by a second LLM (GPT-4) simulating the participant\u2019s continuation. The interviewer agents were instantiated using six contemporary large language models: Claude Sonnet 4, DeepSeek Chat V3 (0324), Gemini 2.5 Pro, GPT-5 Chat, Grok-4, and Qwen3-235B-A22B.